Systems and methods for distributed version reclaim

Abstract

Systems and methods for version reclaiming in a clustered system which avoid attempting to reclaim versions that are still in use by any member of the cluster. Version reclaim is performed efficiently by avoiding duplication of version reclaim effort, avoiding generating version reclaim cleanup transactions that are invalid (i.e., reclaiming versions that are still in use), and minimizing inter-database node communication.

1 . A system for distributed version reclaim in a database cluster comprising a plurality of database nodes, the system comprising:
 a processor; and   a memory storing instructions that, when executed by the processor, configure the system to:   process, by a version reclaim-leader, a release event from a sequence of release events, each release event comprising information about a database node and a version released by the database node;   continuously update, by the version reclaim-leader, a list of database nodes that have released the version until the version has been released by each of the database nodes in the database cluster, provided the version is recognized by the version reclaim leader and provided the version is not reclaimed;   generate, by the version reclaim-leader, one or more cleanup transactions for the version;   commit, by the version reclaim-leader, the one or more cleanup transactions to a transaction log;   continuously replicate, by each database node, the one or more cleanup transactions from the transaction log; and   apply, by each database node, the one or more cleanup transactions for the version.   
     
     
         2 . The system of  claim 1 , wherein when the version is not initially recognized by the version reclaim-leader during processing of the release event, the system is further configured to:
 obtain, by the version reclaim-leader from the release event, a transaction ID associated with the release event; and   replicate, by the version reclaim-leader, a state of the database node based on sequentially applying all transactions preceding the transaction ID, thereby enabling recognition of the version by the version reclaim-leader.   
     
     
         3 . The system of  claim 1 , wherein when releasing the version by the database node, the system is further configured to:
 decrease, by the database node, a reference count for the version until the reference count reaches a minimum threshold.   
     
     
         4 . The system of  claim 1 , wherein the release event comprises a version number of the version, and a database node identifier of the database node. 
     
     
         5 . A non-transitory computer-readable storage medium for distributed version reclaim in a database cluster comprising a plurality of database nodes, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:
 process, by a version reclaim-leader, a release event from a sequence of release events, each release event comprising information about a database node and a version released by the database node;   continuously update, by the version reclaim-leader, a list of database nodes that have released the version until the version has been released by each of the database nodes in the database cluster, provided the version is recognized by the version reclaim leader and provided the version is not reclaimed;   generate, by the version reclaim-leader, one or more cleanup transactions for the version;   commit, by the version reclaim-leader, the one or more cleanup transactions to a transaction log;   continuously replicate, by each database node, the one or more cleanup transactions from the transaction log; and   apply, by each database node, the one or more cleanup transactions for the version.   
     
     
         6 . The computer-readable storage medium of  claim 5 , wherein when the version is not initially recognized by the version reclaim-leader during processing of the release event, the computer is further configured to:
 obtain, by the version reclaim-leader from the release event, a transaction ID associated with the release event; and   replicate, by the version reclaim-leader, a state of the database node based on sequentially applying all transactions preceding the transaction ID, thereby enabling recognition of the version by the version reclaim-leader.   
     
     
         7 . The non-transitory computer-readable storage medium of  claim 5 , wherein when releasing the version by the database node, the computer is further configured to:
 decrease, by the database node, a reference count for the version until the reference count reaches a minimum threshold.   
     
     
         8 . The non-transitory computer-readable storage medium of  claim 5 , wherein the release event comprises a version number of the version, and a database node identifier of the database node. 
     
     
         9 . A computer-implemented method for distributed version reclaim in a database cluster comprising a plurality of database nodes, the method comprising:
 processing, by a version reclaim-leader, a release event from a sequence of release events, each release event comprising information about a database node and a version released by the database node;   continuously updating, by the version reclaim-leader, a list of database nodes that have released the version until the version has been released by each of the database nodes in the database cluster, provided the version is recognized by the version reclaim leader and provided the version is not reclaimed;   generating, by the version reclaim-leader, one or more cleanup transactions for the version;   committing, by the version reclaim-leader, the one or more cleanup transactions to a transaction log;   continuously replicating, by each database node, the one or more cleanup transactions from the transaction log; and   applying, by each database node, the one or more cleanup transactions for the version.   
     
     
         10 . The computer-implemented method of  claim 9 , wherein releasing the version by the database node comprises:
 decreasing, by the database node, a reference count for the version until the reference count reaches a minimum threshold.   
     
     
         11 . The method of  claim 9 , wherein the release event comprises a version number of the version, and a database node identifier of the database node. 
     
     
         12 . The method of  claim 9 , wherein when the version is not initially recognized by the version reclaim-leader during processing the release event, the method further comprises:
 obtaining, by the version reclaim-leader from the release event, a transaction ID associated with the release event; and   replicating, by the version reclaim-leader, a state of the database node based on sequentially applying all transactions preceding the transaction ID, thereby enabling recognition of the version by the version reclaim-leader.